Small Target Extraction Method for Aerial Remote Sensing Images Based on Singular Value Decomposition
To address the accuracy reduction and miss detection caused by low-rank noise interference in the small target extrac-tion of airborne remote sensing images,a extraction method based on singular value decomposition is proposed.This method utilizes singular value decomposition criteria,combined with unevenly changing singular value eigenvectors,to effectively extract small tar-gets and calculate the energy gain of singular value transformation.Based on this,the covariance matrix of clutter in signal space is constructed to reflect the distribution of signals and the connection between signals.Through the singular value decomposition of the matrix,the impact of calculating the clutter covariance matrix is avoided,accurately reflecting the shape,size,texture,and other in-formation of small targets.Furthermore,the image matrix is decomposed to obtain the intensity information of row and column pix-els,and the image matrix is compressed through the orthogonal matrix decomposition and reconstruction.The image is divided into three parts:dispersed,completely superimposed,and partially superimposed targets,and the energy attenuation factor is calculated to achieve the extraction of small targets.Experimental results show that this technique has high recall rate and accuracy,with a maximum miss detection of 4 for ship-like small targets,validating its precise and efficient extraction.
singular value decompositionaerial remote sensing imagessmall object extractionenergy attenuation factor
万方数据
维普
